[antlr-interest] Preserving Unix end-of-line on Windows

[greg94301 at comcast.net]

I'm rewriting files with mixed Unix and Windows end-of-line conventions on a Windows system using the Java code-generation target and Antlr 3.1.3. I'd like to preserve the eol as they appear in the source files, but it appears that the template processor is converting Unix eol to Windows eol. This is a rewriting application, and it looks as if only lines that pass through "-> template" are affected. 

Can I prevent this? 

A stripped-down driver looks like this: 
ANTLRInputStream input = new ANTLRInputStream(new FileInputStream(args[0])); 
TestLexer lexer = new TestLexer(input); 
CommonTokenStream tokens = new TokenRewriteStream(lexer); 
TestParser parser = new TestParser(tokens); 

parser.prog(); 

System.out.print(tokens.toString().length()); 

The test grammar looks like this: 
grammar Test; 
options { 
output=template; rewrite=true; 
backtrack=true; 
memoize=true; 
k=2; 
} 
prog 
: 
IDENTIFIER* -> template(r={$text}) "<r>" 
; 

IDENTIFIER 
: LETTER LETTER* 
; 

fragment 
LETTER 
: 'A'..'Z' 
; 

WS : (' '|'\r'|'\t'|'\u000C'|'\n') {$channel=HIDDEN;} 

Sample test file with Unix eol: 
A 
B 
C 

turns into 
A^M 
B^M 
C 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.antlr.org/pipermail/antlr-interest/attachments/20090829/e674a426/attachment.html